Horse-Race Journalism: Reporting the Polls in the 1976 Presidential Election

Page created by Joann Patterson
 
CONTINUE READING
Horse-Race Journalism:
                        Reportingthe Polls in the
                        1976 PresidentialElection

                                  C. ANTHONY BROH

             The kindsof data obtainedby publicopinionresearchand disseminated
             in themass mediaseemdesignedmoreto entertain  thanto inform.The
             qualityof theinformation
                                    conveyedseemsnotmuchdifferent    fromthat
             conveyedin thesportspagesor,betteryet,thedailyracingform.
                                                                                MILTON ROKEACH(1968)

             The polls have not been as imaginativeas they could have been in
             elucidatingthe real issues and underlyingforces in the election. There
             has been too much fixationon the horse race.
                                                      -ALBERT H. CANTRIL (1976)

A    ND they'reoff.Comingout of the gate,the first"Polls Shows
[sic] Blacks Decisive for Carter." MidwaythroughAugust,"Poll
Shows Slide in CarterMargin."Into September,"Polls Show Ford
Trailingin Bid forTwo VoterGroupsGOP Needs." At thehalfway
markin thecampaign,"VoterPoll FindsDebate AidedFordand Cut
Carter'sLead." As thecandidatesapproachthefinalmonthof cam-
paigning,"First Time Ford Is Ahead of Carter,"and even more
exciting,"Poll Calls Race Tied." Andin thestretch,"SurveyShows
CarterHoldsLead." Atthewireit's Jimmy    Carterelectedpresident.
  Thatis theway thelead storiesin theNew YorkTimescalled the
Bicentennialrunning  ofJimmy  Carterand GeraldFord.The candidate
maytake a shortlead out of the gate,becomefatiguedand change

Abstract Reportingof public opinion polls conformedto a horse-race image of cam-
paign reportingduringthe 1976 presidentialelection. Journalistsavoided prediction,
reportedsegmentsof the sample, selectivelycompared results,emphasized spectacles,
questioned the validityof polling, made a few mistakes, and ignoredcertain data in
theirreporting.All these activitiesreinforcedthe image of elections as a sportingevent.
  C. AnthonyBroh is an AssistantProfessorof Political Science at Columbia Univer-
sityand a Research Associate in the Instituteof Policy Sciences and Public Affairsat
Duke University.

                     Vol. 44:514-529? 1980by The Trusteesof ColumbiaUniversity
PublicOpinionQuarterly
Published by Elsevier North-Holland,
                                  Inc.                                          0033-362X/80/0044-514/$1.75

                 This content downloaded from 128.220.159.74 on Tue, 27 Jan 2015 12:06:58 PM
                                 All use subject to JSTOR Terms and Conditions
HORSE-RACE JOURNALISM                                                                        515

 stridesat the quarterpole, lose momentum       and slow down on the
back stretch,  or winby a nose because of a mediablitzin thehome
 stretch.The race is excitingto followfrombeginning      to end.
   Forjournalists,thehorse-racemetaphor      providesa framework    for
analysis.A horseis judged not by its absolutespeed or skillbut in
comparisonto the speed of otherhorses,and especiallyby its wins
and losses. Similarly, candidatesare pushedto discussothercandi-
dates;eventsare understood    in a contextofcompetition;   and picking
thewinnerbecomesan important       topic.The race-not thewinner-is
the story.The candidate'simage,personality,        staffrelations,and
strategy  are the mainfociof reporting.
   Furthermore,  withthehorse-race   metaphor  journalistscan generate
interest amongvoters,mostofwhomdon'tgetthechanceto observe
thecandidatesin person.By reporting     and interpretingeventsforthe
uninvolvedelectoratethrough     an easilyunderstoodimage,themass
mediaforgea vitallinkin a democracybetweenthepeopleand their
electedofficials. Thus the seemingtrivialization  of one of America's
greatestdemocraticphenomenaactuallyfostersthe democraticpro-
cess.
   However,the imageof a horserace presentssomeproblems.Im-
portantissues of publicpolicymay go unnoticedif the candidates
agree on a position,and conversely,seemingly       unimportant  issues
may receive undue attentionbecause they fit the horse-race
metaphor.For example,in 1976Jimmy        Carter'scontradictory  utter-
ances about the resignation  of FBI DirectorClarenceKelly and his
off-the-cuff statements in a Playboymagazineinterview      aboutadul-
teryreceivedgreatattention.    These storieswerenotaboutissues of
publicpolicy;journalistsperceivedthemin the framework         of cam-
paignstrategy-the   conductoftherace. However,theycan affectthe
outcome.The ClarenceKelly "issue" may have influenced           voters
who thoughtCarterwas "fuzzyon issues." Similarly,         thePlayboy
interview  mayhave adverselyinfluenced    voterswho thought    of Car-
ter as a moral,Christianleader. But neither"issue" informed        the
votersabout Carter'spublicpolicyconcerns.
   Even theevaluationoftheincumbent     president's recordis distorted
by thehorse-race  metaphor.  In 1976thepresspresentedGeraldFord
as a candidateproneto fallsand spills,an imagesupported      by a film
clip of the presidentslippingon the stairsof Air Force One. When
Forddescribedthepeopleof Poland,Yugoslavia,and Rumaniain the
secondteleviseddebateas "freefromSovietdomination,"          thegaffe
reinforced thephysicalimageofa stumbling,     bumbling president. The
statement  was not consistentwithFord's two-yearrecordin office;
yetthestorywas newsworthy     becauseitfitthehorse-race    metaphor.
  Thus the metaphoris both greatlybeneficialand occasionally

               This content downloaded from 128.220.159.74 on Tue, 27 Jan 2015 12:06:58 PM
                               All use subject to JSTOR Terms and Conditions
516                                                                                          C. ANTHONY BROH

harmful to theelectoralprocess.It providesjournalistswitha mean-
ingfulframework,  but thisveryframework  can inducedistortion.It
can heightencitizeninvolvement,butat thecostofignoring important
issues.
  This articlefocuseson theeffectof thehorse-raceimageon jour-
nalists'reporting.               itexamineshowreporting
                 Morespecifically,                      public
opinionpolls conformedto the horse-raceframework   in the 1976
presidentialcampaign.

                                            the Polls
                                 Interpreting
  Aboveall else,publicopinionpollsshowwhois currently       ahead.A
studyof 500 articlesduringthe 1976presidential    election(Carmody,
1976:21)suggeststhatvirtually  all articlesusingpollsreportedwhich
candidateis ahead. My analysisof65 articlesin theNew YorkTimes,
57 nationaltelevisionnews storiesaboutpolls, and 8 nationalnews
magazinestories(Time and Newsweek)betweenLabor Day and
ElectionDay 1976 concurswith Carmody'sconclusion.The few
storiesthatdid not reportwho is ahead dealt exclusivelywiththe
methodology   of the survey.'
  Withintheformat    of reportingwho is ahead,journalistshave con-
siderableflexibility in theirinterpretation  of poll data. They can
predicttheelectionoutcomeor simplyreportthefindings.       Theycan
reportthe percentagesupportforeach candidateor the difference
betweenthe candidates.They can reportthe data for the entire
electorateor for subcategoriesof it. They can select a point of
comparisonwithwhichto interpret     thepolls. Theycan reportvoter
reactionto spectaclesduringthe campaignratherthanreportthe
currentpopularity  of each candidate.They can reportpoll results
accuratelyor erroneously.  They can questionthe validityof a par-
ticularpoll. Finally,theycan ignorepolls. In the 1976 campaign,
reportersmadeuse ofall thesetechniques.In mostcases, theirusage
emphasizedthe horse-raceimageof the campaign.Let us examine
thesepatternsof poll reporting.

  FORECASTING

            theoutcomeof a campaignmaybe themostnewswor-
  Forecasting
thyaspectof a poll. Thus one mightexpectjournaliststo takeevery
  I Reportsofpollsinthe1976electiongenerally
                                           discussedthesampling error,butfew
quotedtheactualwording ofthequestions(see Carmody,1976).TheNew YorkTimes
and televisionnetworknews focusedexclusivelyon the methodology  of surveyre-
searchin 9 percentand 3 percent,respectively,
                                           of theirstorieson polls.

               This content downloaded from 128.220.159.74 on Tue, 27 Jan 2015 12:06:58 PM
                               All use subject to JSTOR Terms and Conditions
HORSE-RACE JOURNALISM                                                                                  517

Table 1. Referencesto PublicOpinionPolls and thePresidencyin theNew York Timesand
          TelevisionEveningNews fromSeptember1 to November2, 1976

                                                                  New York Times TelevisionNews
                                                                     (N = 65)       (N = 57)
 Forecastingof results in report                                            ( 6%)              (16%)
   Forecasts tie or too close to call                                        5                   14
   Forecasts winner                                                          2                    2
 Figures of presidentialtrial-heatreported                                 (15)                (25)
Subcategoriesof population reported                                        (61)                (56)
   Sample of region, state, county,or city only                             31                  24
   Issue preferencecategoryreported                                         15                  12
   Candidate stylepreferencecategoryreported                                 5                    8
   Social groupingreported                                                  23                  10
   Uncertaingroup of voters reported                                        14                    8
Shiftsin polls reported                                                    (12)                (14)
   Compared with pre-Labor Day                                               5                    2
   Compared withpost-Labor Day                                               3                    2
   Compared withprevious election                                            2                    3
   Compared with unspecifiedtime                                             6                  10
Report on spectacle of campaign                                            (17)                (14)
   Debates                                                                  11                  12
   East European gaffe,Playboy Interview,etc.                                3                    0
   Turnoutof Election Day                                                    3                    2
Distortionof poll results                                                   (5)                 (2)
   Reportedin media                                                          3                    2
   Unreportedin media                                                        2                    0
Ignoringresults                                                             (6)                 (3)
  NOTE:  Each entryis the percentage of articles employinga stated technique for
reportingpublic opinion polls. The percentagesdo not total 100o since one articlemay
have used several techniques. Numbers in italics are the net percentagesof articles in
each category.

opportunity   to make predictions.But forecasting has its perilsfor
journalists.Once theyhave forecastthe results,journalistswould
have very littleleftto say; why worryabout the process if the
outcomeseems clear-cut?Moreover,earlyforecastsmightnegate
citizeninterest  in theprocessitself,makethereporter look foolishif
he is wrong,and raise the specterof journalism'sinfluencing     the
outcomeit has predicted.My analysisof the mediafoundonlytwo
instancesof forecasting   the winner,bothverylate in the campaign.
One weekbeforetheelection,theNew YorkTimescarriedan article
quotingpoliticalscientistGerald Pomper,who stated that Carter
would win with54 percent.Two days beforeElectionDay, ABC
carrieda filmclip of Jimmy    Carterstatingthathe wouldwin,while
theFordWhiteHouse, also reflecting    on thepolls,said therace was
                                   thatdidnotuse an outsidesource
too close to call. The onlyforecasts
camewithinthefinaltwoweeksofthecampagin,andforecasta close
race, too close to call. Thus I foundno instancesof journalists
themselves  forecasting thewinner.In total,theTimesreportedfore-

                 This content downloaded from 128.220.159.74 on Tue, 27 Jan 2015 12:06:58 PM
                                 All use subject to JSTOR Terms and Conditions
518                                                                                           C. ANTHONY BROH

casts of any kind in only 6 percentof the articlesabout public
opinion;eveningtelevisionnews storiesdid so in 16 percent.2

      REPORTING THE FIGURES

  Anothertechniquethatheightens   interestin the horserace itself,
not its winner,involvesselectiveuse of the data on presidential
            Of the 65 New YorkTimes articlesreferring
trial-heats.3                                           to public
opinionpolls, 15 percentreportedthe actual percentagesof each
candidateinthepresidential         The other85 percentpresented
                          trial-heat.
the differencebetweenthe candidates,percentages   fora portionof
the population,the answersto questionsotherthanvote intention,
etc. Televisionnews storieswere only slightlymore directeven
thoughvisual presentation of figureswithaudio overlayis equally
appropriate forthismedium.Only25 percentof thetelevisionnews
storiesreportedthepercentages  favoring
                                      each candidatebothorally
and visually.
  Not reporting the percentagesfor the candidateswas a policy
decisionat theTimes,madein an effortto avoidthehorse-race  image
of the campaign.4Ironicallythe policymay have had the opposite
effect.Sincejournalists
                      couldnotreportwhowas ahead,theyfocused
on otheraspectsof the poll-such as the percentagedifference  be-
tweenthecandidates,increases(or decreases)since an earlierpoll,
or, in one case, the actual percentagesreportedby otherpolling
agencies-all ofwhichemphasizethehorse-race   imagebyfocusing  on
changesin the positionof the horseson the track.

  SUBCATEGORIES         OF THE POPULATION

  Reportersmost oftenuse polls to presentthe preferences    of a
subcategory  of the population.They use five subcategories:a re-
gional,state,county,or citysample;a nationalpoll of issue prefer-
ences; a nationalpoll of preferences on candidatestyle;a social
grouping in the population;and uncertainvoters.
  Regionalpolls. Electionyearsproducemanypolls in smallgeo-
  2 The
        differencebetweenmediaforall reporting    techniqueswas not significantly
         X2 = .783; df = 1; p = n.s.
different:
  3 Trial-heat
             refersto thequestionthatasks howpeoplewouldvote iftheelection
wereheldon thedayoftheinterview.     Since 1936,theGallupOrganizationhas always
asked thisquestionwithonlyslightvariations   in wording.In 1976registeredvoters
were asked: "If the PresidentialElectionwere beingheld today,whichcandidates
wouldyouvotefor-theDemocratic     candidates CarterandMondale,ortheRepublican
candidatesFord and Dole?"
  4 See thepaperby William   Kovach in thisissue.

                This content downloaded from 128.220.159.74 on Tue, 27 Jan 2015 12:06:58 PM
                                All use subject to JSTOR Terms and Conditions
HORSE-RACE JOURNALISM                                                                                     519

Table 2. National and State Polls of PresidentialTrial-HeatsReportedin the New York
     Times and TelevisionEveningNews fromSeptember1 to November2, 1976

                                                           New York Times                     TelevisionNews
State
  No directreferenceto agency                                         8                             5
  Field                                                               3                             3
  BuffaloEveningNews                                                  2                             0
  DardenResearch                                                      2                             0
  ChicagoSun Times                                                    1                             1
  ChicagoTribune                                                      1                             0
  MinneapolisTribune                                                  1                             0
  RutgersUniversity                                                   1                             1
  Wilmington News Journal                                             1                             0
  Other                                                               1                             0
  Columbus,OHa                                                        0                             1
  Milwaukee,Wia                                                       0                              1
    Subtotal                                                         21                             12
National
 New YorkTimes(CBS)                                                  27                             6
  Gallup(Newsweek)                                                   13                            13
  No directreference
                   to agency                                          8                            10
  Caddell                                                             6                             1
  Harris(ABC)                                                         6                            18
  Roper(PBS)                                                          4                             2
  Yankelovich(Time)                                                   3                             0
  Other                                                               3                            10
  ChiltonResearch(AssociatedPress)                                    2                             2
  NielsonRatings                                                      2                             0
  NBC News                                                            0                             3
    Subtotal                                                         74                            65
Grandtotal                                                           95                            77
          AgencynotReported.
  a Polling
  NOTE:The networks emphasizedtheircommissioned
                                              pollsmorethanotherpolls;the
ABC-commissioned HarrispollandtheNBC Newspollwerereported   moreon televi-
sionthanin theTimes. Conversely,
                               theTimes referred
                                               moreoftento itsjointlyspon-
soredpollwithCBS thanto anyotherpoll.

graphicareas. The Field Poll, for example,has a reputationfor
reliablyassessingvoteropinionin California;   theRutgersPoll shares
a similarreputationin New Jersey;and severalnewspaperpolls are
widelyrespected.State polls generateinterestamongnonstateresi-
dentsbyproviding  information aboutpivotalstatessuchas California,
New York,New Jersey,   or Illinois.Thus,moststatepollsreported    by
theTimesor theTV networks     wereconductedin northern     industrial
stateswitha history ofbeingon thewinning     side. Such pollsindicate
thetrendsin keystatesthatmayaffecttheoutcome(see Broh,1981).
Thirty-one percentofthereferences   to pollingin thenewspapers, and
24 percenton televisionwere to state,county,or citypolls.
  Issue preferences.Whenissue-oriented   votersprefer  one candidate

                This content downloaded from 128.220.159.74 on Tue, 27 Jan 2015 12:06:58 PM
                                All use subject to JSTOR Terms and Conditions
520                                                                                       C. ANTHONY BROH

on one issue buttheothercandidateon a secondissue,thefocuson
issuesis particularly
                    suitedto capturing voterinterest.A Times/CBS
poll released on October 15, for example,showed the candidate
preferencesofvoterswithvariousissuepositions.The report(Apple,
1976:B4)states:"One ofthekeyissuesis economic.The Times/CBS
surveyshowedforexample,thatMr. Carteris winningoverwhelm-
inglyamongthosewhociteunemployment       as theirdominantconcern,
as wouldbe expected."
  But the same poll showedthatCarterwas onlyeven withFord
amongthoseconcernedaboutinflation.    ThustheTimesreaderwhois
worriedaboutbothunemployment     andinflation receiveda messageof
a closelycontestedcampaign.
  Accordingto a reportof theNationalCouncilon PublicPolls and
theGannettJournalism  Centerat Northwestern   University,8 percent
of the newspaperarticlesin 1976 dealt withissues in thismanner
(Carmody,1976:21).Thesefindings   concurroughly   withmyanalysis,
whichshows 15 percentof theTimesreportsand 12 percentof the
televisionnewsreportsfocusing  on thissubcategory. Whilenotover-
whelming, thisamountofreportage   on issuessuggeststhatthehorse-
race metaphordoes not lend itselfto reporting   on issues thatare
complexor hotlybut simplistically  contested.
  Preferences on candidate style. For example, the New York
Times/CBS surveyshowedCarterleadingamongpeoplewhosaw him
as a conservative(Reinhold,1976a; 1976b:32).Five percentof the
New YorkTimesarticlesand 8 percentofthetelevisioneveningnews
broadcastsabout public opinionpolls focusedon preferences    for
candidatestyle.Storiesreferred to confidencein the president,
                                                            and
thehonesty,credibility,and characterof bothcandidates.Whilethe
characterof the candidatesis obviouslyof crucialimportance fora
meaningful democraticchoice,thehorse-racemetaphor   runstherisk
of emphasizingbeauty-somehorsesare gorgeousanimals-and ne-
glectingdifferenceson issues of substance.
  Social groups. People of differentreligions,occupations,races,
regions,and ages have varyingcandidateand issue preferences.  In
particular,
          nationalpolls oftenstriveto reporttheopinionsof mem-
bersoftheold New Deal coalition,especiallywhena particulargroup
seemsto be pivotalin theelectionoutcome.For example,Catholic
voterswerepolledafterFordreferred  to East Europeansas freefrom
Soviet domination, and Jews were polled when New York State
appearedcrucialto Carter'ssuccess. Twenty-three   percentof the
Timesarticlesand 10 percentof thetelevisionnews reportsfocused
on thistypeof pollinginformation.
  Uncertainvoters.Voterswhoidentify   withneithertheDemocratic

            This content downloaded from 128.220.159.74 on Tue, 27 Jan 2015 12:06:58 PM
                            All use subject to JSTOR Terms and Conditions
HORSE-RACE JOURNALISM                                                                        521

northeRepublicanparty-andwhoare therefore    unpredictable-may
be seen as determiningthe       outcome.
                         election         Thus on October15,the
New YorkTimesfront-page   story(Apple, 1976:A1) had thefollowing
headlines:"Carter,Focusingon Ford Record,Gains AmongInde-
pendentsin Poll." Similarly,"undecided" voters,who are highly
volatileandcan decideforeithercandidate,are importantsubjectsfor
reporters(see DeVries and Tarrance,1972). The New YorkTimes
began its predebatecampaignstorywith the followingheadline:
"Large Groupof UndecidedVotersFound Lookingto Debates for
Aid: StudyAlso Shows Much of Supportfor Carterand Ford Is
Subjectto Quick Change" (Reinhold,1976a).
  Like undecidedvotersor independents,  peoplewho supportthird-
partycandidatesmay affectthe resultsand therefore  receivegreat
attentionfromhorse-race-conscious journalists.Eugene McCarthy
was portrayed as a candidatewithsupportamongyoungvoterswho
recalledhis anti-Vietnam  War positioneightyears earlier.These
votersweredisproportionally Democratsand thuscouldconceivably
swingto supportCarter.On the otherhand,LesterMaddox could
siphonsupportaway fromfellowGeorgianJimmy       Carter.Thus,the
mainemphasisof a September9 NBC News reportof a Harrispoll
(showingCarterin the lead) was the effectof the McCarthyvote,
whichmade the outcomeof the electiondifficult  to forecast.Simi-
larly,Time(1976b)reported on October11that"The Race Turnsinto
a Dead Heat" withthefollowing analysis:"The standoffturnedup by
the surveyresultedfromaskingthosepolledto choose betweenthe
two maincandidates.WhentheYankelovichanalystsfiguredin the
effectof the minorcandidates,LesterMaddox and Eugene McCar-
thy,FordpulledaheadofCarterby42% to 40%." Fourteenpercentof
theNew YorkTimesarticlesand 8 percentof thetelevisionevening
news broadcastsaboutpublicopinionpolls focusedon thisformof
uncertainty,thusenhancingvoterinterestin the electoralprocess.

  SHIFTS IN POPULARITY

  Another            techniqueforenhancing
          journalistic                     interestthrough  polls
is comparingcurrentand earlierpoll data. Journalists can choose
comparisons in orderto describethecandidatesas close or farapart,
butusuallytheirreportsimplyconstantly changing opinion,so thata
candidatecan seemto wintheracewitha suddenburstofpopularity.
            mustdecidewhatto comparetheirdata to: earlierpolls
  Journalists
by the same agency,otherpolls takenat the same time,polls in
previousyearsat the same pointin the campaign,and the like. An
exampleof a reporter'suse of thisdiscretionto createexcitement

               This content downloaded from 128.220.159.74 on Tue, 27 Jan 2015 12:06:58 PM
                               All use subject to JSTOR Terms and Conditions
522                                                                                          C. ANTHONY BROH

turnedup in an NBC interview    withGeorgeGallupon October1,
1976,whichconcludedthatcurrent    trendsin Carter'slead overFord
would producean election   too close  to call (VanderbiltTelevision
News Archive,  1976).This statement  emphasized  theshiftin opinion
whileignoring theactual11percentdifference   betweenthecandidates
in Gallup'spoll.
  A problemwithcomparing    pollsat one timeto pollsat anothertime
                   of samplingerror.Since mostcommercial
is the interpretation                                         polls
havea samplesize of 1,000to 1,500voters,thesampling    errorcan be
expectedto be about 3 or 4 percent.The actualfigurewilldepend
uponthe samplingtechniquesused by theagencyand chancevaria-
tion(Weisbergand Bowen, 1977).In recentyearsmostnewspaper
accountsof polls reportsamplingerror(Carmody,1976)and sample
size; however,thepressrarelyconsidersthisfactorininterpreting  the
results.5For example,severalarticlesreporteda shiftin popularity
measuredby subtractingpopularityat t2frompopularityat t,. Thus a
changefrom58 percentto 54 percentwouldhavebeenreported     as a 4
percentdecrease. However,the 4 percentmay not signifyattitude
changesincea popularity ratingof56 percent,forexample,at botht1
andt2is withinthesampling  errorofbothfigures.Thuson October1
theTimesreportshowingan 18percentdifference     betweenCarterand
Ford at t1,and an 8 percentdifference  at t2 may,or maynot,have
reflecteda netattitudechangeof 10percent.Nevertheless,  theTimes
reportedthe figuresas attitudechange. Similarlythe "Fluttering,
StutteringPolls" describedby Timemagazine(1976b)in a postelec-
tionarticlecouldhave been due to a neglectin considering sampling
error.
  A problemwithcomparing     polls of one agencywiththe polls of
anotheragencyis the variationin undecidedvoters,whichchanges
for several reasons. Some votersdecide how to vote late in the
campaign;samplingerrorproducesfluctuations     in thepercentageof
undecidedvoters;and thepollingagenciesgiveinterviewers   different
instructionsaboutpressingrespondents   foran answer.Speakingbe-
foretheAssociatedPressManagingEditorsConference,      BurnsRoper
noted,"There is also the problemof decidingwhetherto forcethe
undecidedvotersto [make]a choice,and,ifso, how and howforce-
fullyand whetherto allocatethe remaining    undecidedvotersto the
candidates"(quotedin Carmody,1976).Clearly,thenumberofunde-

  S The American AssociationforPublicOpinionResearchhas publishedminimum
standardsof disclosurefornews releasesaboutpublicopinionpolls. It urgesnews
                          information
mediato includethefollowing          whenpreparingfinalcopyforpublicationor
broadcast:sponsor,questionwording,population,samplesize, subcategoriesof re-
sponse,typeof interview,and timeof survey.

               This content downloaded from 128.220.159.74 on Tue, 27 Jan 2015 12:06:58 PM
                               All use subject to JSTOR Terms and Conditions
HORSE-RACE JOURNALISM                                                                        523

cided voterschangedfrompoll to poll in the 1976campaign.6  These
changesmayhave comeaboutbecauseofrandomvariationor biased
resultsdue to interviewtechniques.Thatis, theundecidedvotersmay
disproportionatelyfavorone candidateoveranother.7  Mostimportant
to our discussion,thebias in one samplemaydiffer  fromthebias in
another.Journalists willreporttheresulting           in candidate
                                           differences
supportas attitudechange,and thusmakea campaignappearmuch
morevolatilethanit reallyis.
  Furthermore,  in 1976the selectionof a date forcomparisonpro-
duced different imagesof the race. By comparingthe candidates'
popularityduringthecampaignto popularity   beforeLabor Day,jour-
nalistscouldpresenta catch-upimageoftheFordcampaign.By com-
paringthe candidates'popularity  duringthe campaignto popularity
afterLaborDay,journalists  couldpresenta nose-to-noseimageofthe
campaign.A thirdpossibilitywas to compareCarter'slead in the
polls to othercandidates'popularityin previousyears. Franklin
Roosevelt'sand RichardNixon'slandslidevictoriesin 1936and 1972,
respectively,were typicalpointsof comparison,and were used to
emphasizeCarter'sslippagein popularity.   Table 1 reportsthe per-
centageofarticlesintheNew YorkTimesandtelevision    eveningnews
storiesaboutpolls whichreporteda shiftin publicopinionin com-
parisonto some priorpoint.

  REPORTING CAMPAIGN EVENTS

  Another  reportingtechniqueforgeneratingexcitement is to empha-
size spectaclesor specialeventsduringthecampaign.The New York
Timesdid thisin 17 percentof its reports,televisionin 14 percent.
The televiseddebates,forexample,showedthe candidatesin face-
to-facecompetition, permittingthe armchairjockey to evaluatethe
race as itheadedtowardthefinishline.Furthermore,  sincecampaign
staffsprovidepredebateplansand strategy, and newscommentators
emphasizeeach candidate'sviability forofficebased on his perform-
ance,themostfeeblecandidatecan appearmorecapable-and thusa
close contenderin the election.
  By reporting  figureson publicperceptionsof the debate winner
whileignoring  figureson overallpublicpreferences, theNew York
  6 BetweenLaborDay andElectionDay theTimes reported  theresultsof14different
nationalpolls. The numberof votersreportedas notfavoringeitherCarteror Ford
variedfrom18percentto 7 percent;theseextremeswerereportedon thesameday(see
Broh,1979).
  7People who respond"I don'tknow" to a surveyquestionare disproportionally
nonwhite,low-educated,low-income, and noninvolved
                                                (see Francisand Busch,1975;
Converse,1976-1977).

               This content downloaded from 128.220.159.74 on Tue, 27 Jan 2015 12:06:58 PM
                               All use subject to JSTOR Terms and Conditions
524                                                                                        C. ANTHONY BROH

 Times heightened  the citizen'sinvolvement  withthe spectacle.In
 addition,the media emphasizedthe degreeof competition,     not the
 victory.On September  24, theday afterthefirstCarter-Ford  debate,
 NBC reportedtheAssociatedPresspoll callingthedebateslightly     in
 favorofPresidentFord,butemphasizedthat24 percentbelievedthe
 debateto be a draw.Two dayslaterCBS placedsimilaremphasison
 the39 percentin itspoll who had no opinionaboutthewinnerofthe
 debatesor thoughtthe debate was a draw. To the extentthatthe
 debateschangedvoters'mindsaboutthecandidates,theresultswere
 "closingthe gap on Mr. Carter"(VanderbiltTelevisionNews Ar-
chive,1976:1997).
   By heightening voterinterest,thisselectiveemphasisin reporting
pollsis a valuableaspectof themedia'srole in thedemocraticpro-
cess. Butthemediasometimes    go beyondwhata democratic     theorist
mightwish by encouragingvotersto focus on events that affect
campaignstrategy   but thatmay have littleimpacton a president's
behavior.These media-created   spectaclesare thenused to analyze
trendsor shiftsamongvoters.For example,in an interviewwith
Playboymagazine,Jimmy     Carterdiscussedhis religionand "lust for
women." ObviouslyCarter'ssexual desiresor the factthathe was
interviewed in a "sex magazine"wouldnotaffectpolicypositionsor
issuesafterthecampaign.However,thesequestionsmayhave influ-
encedmanyvotersafterthemediasingledout thePlayboyinterview
as an important eventanda cause ofincreasing  electoralcompetition.
Changesin thepolls,accordingto reporters,  resultedfromthispoliti-
cal miscalculation.Similarly,GeraldForderredin theseconddebate
whenhe said: "I don't believe,Mr. Frankel,thatthe Yugoslavians
considerthemselves   dominated by the SovietUnion.I don'tbelieve
thatthe Poles considerthemselvesdominatedby the SovietUnion.
Each of these countriesis independent   or autonomous"(Congres-
sional Quarterly Almanac, 1976:930).Afterthegaffe,publicopinion
pollsreported thatCatholicsof East Europeanancestrywerereturn-
ingto theDemocraticranks.The highconcentration      of Catholicsin
Detroit,Chicago, Cleveland,and Milwaukeecould have affected
Ford's chances in Michigan,Illinois,Ohio, and Wisconsin-states
crucialto GOP strategy.  By focusingon event-linked changesin the
polls,themediaencouragedcitizensto weighstatements      in termsof
theireffecton the campaign,not on UnitedStatesforeignpolicy.
  Eventsoccurring   duringthecampaignaffecta candidate'ssupport
and thushis positionin the race. Each mistake,each appearance,
each stumbleis important  because it affectsthecandidate'sstrategy
and mighteven cause his campaignto falterfora moment.By em-
phasizingthese events reportersfocus on the volatilityof public

             This content downloaded from 128.220.159.74 on Tue, 27 Jan 2015 12:06:58 PM
                             All use subject to JSTOR Terms and Conditions
HORSE-RACE JOURNALISM                                                                        525

opinion,theunpredictability
                         oftherace,and thesignificance
                                                    ofcam-
paignstatementsratherthanactualbehavior.

  INACCURACIES IN REPORTING

  In a fewinstances,minorinaccuraciesoccurredin thereporting      of
polls,and alwaysin a mannerthatindicateda close race. For exam-
ple,on September   2, David Brinkleyreported thelatestGalluppollas
givingCarter52 percentand Ford 37 percent.The Gallup Organ-
ization'spostelectionreportshowedthe resultsof the same poll as
54-36 (Gallup,1976:13),a 3 percentwiderdifference      thanthe NBC
spread.Timemagazinereported     a YankelovichpolloflateSeptember
as a tie, 43 percentto 43 percent(Time,1976b:16). A monthlater,
Time's(1976a:17)retrospective  analysisshowedthatFordactuallyled
inthepoll,43 percentto 41 percent.Although   thedifference  was only
in allocationof the McCarthyvote, the reportconfoundedthe in-
terpretation of a 4 percentlead takenby Carterin the following
month'spoll,and negatedthecontention     oftheSeptemberstorythat
the campaignwas a "dead heat." Similarly,the New YorkTimes
exaggerated  theeffectofthefirstdebatein itsreportof a Galluppoll
showingan 8 percentdifference    betweenthe trial-heat  popularityof
Carterand Ford.The poll,takenSeptember     24 to 27 andthefirstafter
thedebates,was highlyinfluential  in producing theillusionthatFord
was "catchingup." The October1 New YorkTimes(1976a:8)story
hadthefollowing  lead: "CarterMarginin GallupPoll Is Cutto 50-42.
The latestGallupPoll on thePresidential   contestshowsthatJimmy
Carter'slead overPresidentFordhas beenreducedfrom18 percent-
age pointsto 8." Gallup's(1976:13)postelection   analysisreportsthe
poll resultsas 51 to 40 percent.
  Withno attempt  to deceiveor misleadvoters,thecumulative    effect
oftheseindividually trivialinaccuraciesis nonetheless to heightenthe
senseof competition  and to sharpentheimageof a horserace. They
are perhapsmostimportant     as a demonstration  of the unconscious
powerofthedriveto maintain    voter-andreaderor viewer-interest.

  IGNORING RELEVANT DATA

  Stillanothertechniquefordealingwithpoll data is to ignoreor
downplaythem,especiallyifthepoll showsthecandidatesapart.For
example,theNew YorkTimesand CBS commissioned      severalpolls
duringthe 1976presidential
                         campaign,one of whichwas conducted
betweenOctober8 and 12 and releasedto thepublicon October15.
The GallupOrganizationalso conducteda poll duringthe same pe-

               This content downloaded from 128.220.159.74 on Tue, 27 Jan 2015 12:06:58 PM
                               All use subject to JSTOR Terms and Conditions
526                                                                                           C. ANTHONY BROH

riod, October 8 throughOctober 11. The Times (Apple, 1976) empha-
sized the six-pointdifferencereportedby Gallup and played down its
own poll which made the race seem slightlyless competitive.The
fourthparagraphof the page-one storyread as follows:

                                6 Point Edge for Carter
  A new nationalsurveyby the Galluporganization,  the resultsof which
weredisclosedyesterday, gave Mr. Cartertheedge by 6 percentage points,
48 to 42. A simultaneous
                       nationalpoll by PatrickCaddellforMr. Carterput
his marginat 8 points.The Times/CBSfindings,  whichwerebased on tele-
phoneinterviews with1,761registered votersbetweenOctober8 and 12,fell
intothe same generalrange.

The lead sentence suggeststhatthe 6 percentedge is "correct," while
other percentages are either biased-since Caddell worked for
 Carter-or close enoughto be "the same." However, theTimes's own
poll showed 44 percent for Carter and 37 percent for Ford-a 7-
percentdifference.8  While 7 percentis clearlywithinthe range of the
Gallup poll difference,theTimes reportedand emphasized the smaller
figureof an outside polling agency. Moreover, the referenceto the
Times poll which emphasized its methodology (" 1,761 registered
voters between October 8 and 12") had the effectof bolsteringthe
reliabilityof the Gallup poll, whose sample size it did not report.
   NBC News explicitlyignoredthe poll data in its September16 news
report(VanderbiltTelevision News Archive, 1976:1941). JohnChan-
cellor announcedthatvice presidentialcandidate WalterMondale was
campaigningaround the country,andthen stated, "Despite public
opinionpolls, Carterand Mondale may be in a tightrace." The story
that followed showed Mondale on the campaign trailand Democrats
and labor leaders saying the race was too close for comfortin their
states of Ohio, California,Michigan, Indiana, New Jersey,and Il-
linois. The storyimpliesthatlocal leaders "know" how close the race
really is while the national polls are misleading.This kind of report
did not appear later in the campaign when the nationalpolls showed
the candidates to be very close.

                             Conclusionsand Discussion
  The horse-race metaphor of campaigns is not new. The Boston
Journal in 1888 proclaimedthata "dark horse" is unlikelyto emerge
fromthe campaign. In 1924, the New York Times ran the following
  8 Information received by the authorin a telephoneconversationwithDwight Mor-
ris, March 20, 1980.

                This content downloaded from 128.220.159.74 on Tue, 27 Jan 2015 12:06:58 PM
                                All use subject to JSTOR Terms and Conditions
HORSE-RACE JOURNALISM                                                                        527

 headlineabout the Democraticconvention:"With 21 Candidates, It Is
 Now The Field AgainstM'Adoo" (Keylin and Nelson, 1976). What is
 noteworthyis that polling technologyof computer-drawnsamples,
 random digit dialing, and carefullyconstructedquestions has con-
 formedso easily to the traditionalreportingtechniquesofjournalists.
   Viewing a presidentialcampaign as a horse race, then, has a long
 tradition.A democratic theoristmightdecry this rathercrude con-
 nectionbetween a sporting,indeed a betting,event, and the ideals of
democracy. Where in all of this is Rousseau's informedand commit-
ted participantin the events that shape his life,Jefferson'sfreedom-
loving farmer,de Tocqueville's equality-lovingentrepreneur,V. 0.
Key's rationalvoter? But it is one thingto uphold democraticideals;
it is a verydifferent thingto condemnthe media forusing a sporting
image to frametheiranalyses of elections.
   The horse-race metaphorin fact has several valuable functions.
First, and most important,it enhances the public's interest in a
process that could seem remote,mysterious,and boring. If citizens
are more attractedto sports than to politics, why not use sports to
teach them about politics? Because it makes the electoral process
appear as excitingand competitiveas possible, the horse-raceimage
maintainsa vital link between the mass of people and a very few
elected officials.Second, the horse-race image induces reportersto
focus on polls that hold great interestfor special groups of voters.
Reportingthe views of Californians,or Catholics, or proponentsof
gun controlmay not give us much sense of how our candidate is doing
overall, but it is of vital importance to westerners, the religious
community,and gun owners. Finally, the horse-raceimage mitigates
the danger inherentin the very existence of the mass media-that
reporterscould have much more influencethan they or others want
by settlingon a winnerearly in the process and turningthe campaign
into a self-fulfilling
                    prophecy. The horse-race image encourages re-
portersto emphasize competitionratherthan to forecast results.
   However, we mustconclude witha note of caution. My analysis of
the 1976 election shows two problems in the use of the horse-race
metaphor.First,reporterswho seek a themeto make a storyexciting
may inadvertently  distortpolls to the pointof nonsense. Burns Roper
(1976) illustratedthe discrepanciesin the use of his poll in a satirical
letterto the editor of the New York Times:
To theEditor:
... It had notoccurredto meto combinefromour"instant"debatepollthe
31 percentwhothought  GovernorCarterwonwiththe30 percentwhofeltit
was a drawand conclude". . . a totalof 61 percentof thosepolledthought
thatthe debatewas a drawor a Cartervictory."Nor was I perspicacious

               This content downloaded from 128.220.159.74 on Tue, 27 Jan 2015 12:06:58 PM
                               All use subject to JSTOR Terms and Conditions
528                                                                                         C. ANTHONY BROH

enoughto recognizethat"in politicalterms,"thatwas an excellentresultfor
thenomineeofthestronger party.[However,I becamemoreconfusedwhenI
read]thatsincethePresident'sstrength priorto thedebatewas onlyabout33
percent(moreor less, dependingon whichpoll you read),thedebatewas a
victoryforthe Presidentsince his win-plus-draw figurewas 61 percentac-
cordingto theAssociatedPresspoll-roughlytwicethepercentage   preferring
him prior to the debate.       . .   . I had looked simplisticallyat our 39 percent
Ford,31 percentCarter,30 percentdrawfiguresand concludedthatPresi-
dentFordhas a smalledge.It is nowobviousto me thatitwas eithera clear
victoryforGovernorCarteror an overwhelming  winforPresidentFord.

To Roper, scientificallydrawn samples are more credible thanjour-
nalists' impressions-except when analystsuse the polls to see what-
ever they like. Polls are as credible as their users; the search for
excitementcarries withinit the danger of distortion.
   Finally, the horse-race image can encourage voters to focus on
exciting, but ultimatelyirrelevantaspects of a campaign. Just as
bettors on a race may be misled by the beauty of the horses-
champion thoroughbredsare often the least showy-voters may be
misled by the triviaof a campaign. Pseudo-eventsthatare created by
the media, and whose effectson poll resultsare thenanalyzed by the
same media forclues to the dynamicsof a campaign,do the voters a
disservice. Conversely,issues which are complex, not hotlydebated,
but more significant,may be passed over in the search forexcitement.
  In sum, the horse-race image frames the analysis of polls in an
election with generallybeneficialand sometimesdeleterious effects.
By enhancingthe excitementof the campaign, the media draw the
electoratecloser to theirelected representatives,withonly occasional
costs to the integrity
                     of the poll data or the ideal democraticprocess.

                                         References
Apple,R. W.
 1976 "Carter,focusingon Ford record,gains amongindependents
                                                           in
        poll." New YorkTimes, Oct. 15, pp. Alff.
Broh,C. A.
  1979 PublicOpinion,Pollingand thePress:The Race forthePresidency.
        Durham,N.C.: Duke University   Instituteof Policy Sciences and
        PublicAffairsWorkingPaper.
  1981 "Whetherbellwethers    or weather-jars
                                            predictelections."Western
        PoliticalQuarterly
                         (forthcoming).
Cantril,A. H.
  1976 "The press and the pollster."Annalsof the Academyof Political
        and Social Science 427:45-52.
Carmody,D.
  1976 "Poll takersvoice some reservations  in analysisof coverageby
        newspapersof campaignpolls." New YorkTimes,Nov. 13, p. 21.

              This content downloaded from 128.220.159.74 on Tue, 27 Jan 2015 12:06:58 PM
                              All use subject to JSTOR Terms and Conditions
HORSE-RACE JOURNALISM                                                                        529

CongressionalQuarterlyAlmanac
  1976 Washington,                   Quarterly.
                   D.C.: Congressional
Converse,JeanM.
  1976- "Predictingno opinionin the polls." Public OpinionQuarterly
  1977 40:515-30.
DeVries,W., and L. Tarrance
                          GrandRapids:WilliamB. Eerdmans.
  1972 The Ticket-Splitter.
Francis,Joe D., and LawrenceBusch
  1975 "Whatwe nowknowabout'I don'tknows'." PublicOpinionQuar-
       terly39:207-18.
Gallup,G. H.
  1976 "Presidentialtrialheats,1976." The GallupOpinionIndex. Report
       No. 137:13.
Keylin,Arleen,and Eve Nelson
  1976 If Elected     . . .   New York: Arno Press.
New York Times
  1976a"Cartermarginin Galluppoll is cut to 50-42." Oct. 1, p. 8.
  1976b"Poll showsslidein Cartermargin."Aug. 25, p. 1.
Reinhold,R.
  1976a "Large groupofundecidedvotersfoundlookingto debateforaid."
       New YorkTimes,Sept. 22, p. 24.
  1976b"SurveyshowsCarterholdslead becausehe is seen as moderate."
       New York Times,Oct. 19,p. 32.
Rokeach,M.
  1968 "The role of values in publicopinionresearch."Public Opinion
        Quarterly32:547-59.
Roper,Bums
  1976 "Letters." New YorkTimes,Oct. 7, p. 46.
Time
  1976a"Cartertakesa narrowlead." Nov. 1, pp. 17-18.
  1976b"The race turnsintoa dead heat." Oct. 1, pp. 16-17.
VanderbiltTelevisionNews Archive
  1976 Televisionnews index and abstracts.Nashville:JointUniversity
       Libraries.
Weisberg,HerbertF., and BruceD. Bowen
  1977 An Introductionto SurveyResearchand Data Analysis.San Fran-
       cisco: W. H. Freeman.

               This content downloaded from 128.220.159.74 on Tue, 27 Jan 2015 12:06:58 PM
                               All use subject to JSTOR Terms and Conditions
American Association for Public Opinion Research

Horse-Race Journalism: Reporting the Polls in the 1976 Presidential Election
Author(s): C. Anthony Broh
Source: The Public Opinion Quarterly, Vol. 44, No. 4, Polls and the News Media: A Symposium
(Winter, 1980), pp. 514-529
Published by: Oxford University Press on behalf of the American Association for Public Opinion
Research
Stable URL: http://www.jstor.org/stable/2748469 .
Accessed: 27/01/2015 12:06

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at .
http://www.jstor.org/page/info/about/policies/terms.jsp

.
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of
content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms
of scholarship. For more information about JSTOR, please contact support@jstor.org.

.

                American Association for Public Opinion Research and Oxford University Press are collaborating with JSTOR
                to digitize, preserve and extend access to The Public Opinion Quarterly.

http://www.jstor.org

                             This content downloaded from 128.220.159.74 on Tue, 27 Jan 2015 12:06:58 PM
                                             All use subject to JSTOR Terms and Conditions
You can also read